Exercise 14 of the Udacity class on Machine Learning workflows. The objective of this exercise is to bring the lesson's contents together in a complete ML pipeline that produces a trained Random Forest model.
The dataset used in this exercise is a modified version of the original songs dataset: here
We use a Random Forest to perform songs' genre classification (NLP).
The model was evaluated using a test set from the dataset, with an AUC (Area Under the Curve) of 0.95326. The following figures show the feature importance and the confusion matrix computed during the evaluation step:
You can use this pipeline at version 1.0.0 with mlflow and w&b (you need to be logged in: wandb login):
mlflow run -v 1.0.0 git@github.com:AMergy/genre_classification.git
for instance:
mlflow run https://github.com/AMergy/genre_classification.git \
-v 1.0.0 \
-P hydra_options="main.project_name=remote_execution"